A novel graph approach to resolve challenging and medically relevant regions in the human genome

The National Institutes of Health National Human Genome Research Institute awarded Baylor College of Medicine $719,909 on March 13, 2026, under the Human Genome Research program (CFDA 93.172) to develop and apply graph genome methods for characterizing medically relevant complex genomic regions.

Baylor will construct a graph genome from approximately 15,000 long-read genome assemblies sourced from existing NIH programs (All of Us, Human Pangenome Reference Consortium) and 50 locally generated long-read assemblies from a cardiovascular risk study (HeartCare). The work focuses initially on the LPA gene and its complex hypervariable Kringle IV Type 2 region, which harbors variants potentially impacting cardiovascular disease risk. Newly identified LPA variants will be analyzed for association with deep phenotypic measurements across 30,000 whole-genome sequences. The methods will then be scaled to approximately 1,000 challenging but medically significant genes, including the 73 genes on the American College of Medical Genetics secondary findings list, 132 HLA genes, and 395 additional clinically relevant genes. The deliverable is a comprehensive annotated variant catalog accessible to non-specialists. Baylor will collaborate with individual investigators from ARIC, SOL, All of Us, GREGOR, and TOPMed cohorts to validate pathogenicity of newly discovered variants.

Performance occurs in Texas. The period of performance runs from the award date through February 28, 2030. This is a Project Grant, an assistance type used for targeted research projects.

Abstract

Despite improved human genome references, like the T2T and human pangenome references, limitations of short-read sequencing and inadequate tools preclude routine characterization of medically relevant complex regions like LPA, HLA, and GBA. This includes 'graph genome' methods, which have not yet been effectively applied to the characterization of clinically important, complex genes. We will develop a novel user-friendly and intuitive local graph approach that will be applied to resolve these critical genomic regions and yet accessible to non-specialists who wish to study complex regions of the human genome. A graph genome will be built from ~15,000 long-read genome assemblies from existing data from NIH programs (all of US, hprc) and 50 locally generated long-read assemblies from a cardiovascular risk study (heartcare) and then benchmarked and validated against existing references and methods. This graph will allow US to utilize existing short-read data to deepen our insights into the variation in LPA and its complex hypervariable kringle IV type 2 (KIV-2) region, where additional variants potentially impact cardiovascular disease (CVD) risk. Newly assessed LPA variants will be analyzed for association with deep phenotypic measurements across 30,000 WGS and positive findings will greatly impact the application of LPA testing outside european populations. We will also apply our graph genome methods across larger data sets from topmed and all of US to assess around 1000 challenging but medically significant genes, including the american college of medical genetics secondary findings gene list (73 genes), 132 HLA genes, and a further 395 clinically relevant genes, providing the most comprehensive annotated variant catalog of its kind. We will work with individual investigators (aric, SOL, all of US, gregor, topmed) to further validate the pathogenicity of newly identified variants, across multiple genetic diseases. We will also work with the topmed IRC to include these variants in imputation servers (bravo). Overall, this proposal will enable the use of graph genomes at scale and demonstrate their utility, impacting both the assessment of CVD risk across different populations and providing new information for multiple other diseases.